| In the Internet+era,due to the popularity of smartphones and information systems,companies have accumulated a large amount of operational data,such as customers,orders,products and other information.The analysis of data-based customer behavior and the subsequent business operation management optimiza-tion are the research hotspots in the business community and academia research in recent years.Although some studies have considered the identification of cus-tomer behavior patterns on the basis of data or further optimize the operational decisions of enterprises,there are still many problems to be solved in practical applications.Based on the existing research,this paper starts from the actual operation data of enterprises,finds problems in different enterprise scenarios,and pre-dicts customers' purchase behavior and optimizes enterprise operation decisions through mathematical modeling and solution.The relationship management be-tween enterprises and customers is generally divided into two scenarios,one is the service subscription research under the contract scenario,and the other is the product purchase research under the non-contract scenario.This paper combines enterprise'consulting experience and field research,uses statistical model method such as survival analysis and machine learning to model the customer behavior in different scenarios on the basis of the actual data,trains model through the training set data.The model is tested by validation set data,and the new model is proposed and compared with other benchmark model analysis,thus effectively help enterprises to further understand their customer base.In addition,on the basis of recognizing customer purchase behavior pattern,we further expand the corresponding inventory optimization model,and apply the results of the theo-retical model to the case analysis,to help enterprises improve the corresponding decision analysis ca.pabilit,ies.The main work and research results of this paper are as follows:(1)For the customer service subscription forecasting problem in the contract scenario,through the analysis of enterprise data,we found that customer churn behavior is related to the effect of operators,subscription channels and customer individual characteristics.Based on this,this paper proposes a new two-stage bayesian network predictive model to predict customer churn behavior.In the case analysis,the proposed model is tested using the actual data of the enterprise,which also verifies that the model is helpful in predicting the effectiveness of the customer churn behavior.(2)For the prediction of customer product purchase in the non-contract sce-nario,we found that the traditional model assumptions were not applicable and only used part of the information,so we built a more extensive prediction model on the basis of enterprise data analysis.The model includes the prediction of customer loss and the prediction of customer repurchase,and extends the hetero-geneity of customer from the traditional distribution description to the regression fitting with customer specific attributes.The proposed model is tested with the actual data of enterprises in case analysis,and the validity of the model in helping enterprises predict customers' repeated purchase behavior is verified.(3)This paper studies the online promotional gift giving strategy of an en-terprise.Firstly,empirical study is conducted on enterprise data to examine the influence of different gift giving strategies on customers' purchasing behaviors.We found that online promotional gifts can increase the number of repeat purchases and shorten the average purchase interval,but do not stimulate customers to in-crease the amount of consumption.Based on the above conclusions,we propose a dynamic and multi-gift inventory model,which considers the problem of gift in-ventory optimization with or without budget constraints,solves the corresponding optimal gift allocation strategy,and guides enterprises to optimize their profits through different gift allocation strategies through case analysis. |
PDF Full Text Request